Hot take: the vibe coder space has too many builders and most of them don’t know how to market their product.

This is seriously terrifying.

How to build a unicorn startup: Step 1: lock yourself in your room Step 2: stay locked-in until you are profitable Step 3: sell to Elon for $60b

Story time about my generational fumble with Cursor and how I lost many millions: (If you just care about the money takeaway and not the backstory, skip to the last paragraph.) I was one of the earliest @cursor users back when it was just a chat in a sidebar that worked on an 8000 context window of GPT-3.5 with a vector database. I could feel right away that the models were only going to improve and that someday programming as we know it would be finished. The days I was fantasizing about 3 years ago are the days I live in right now, where I can clone any app with one prompt. It all feels so natural as the years rolled by, but feeling that insight back then was so compelling to me that I abandoned everything else I was doing to work on coding agents. I interviewed at Cursor and it didn't work out. But I still loved the problem so much that I kept exploring coding agents. There was only one open source coding agent from @sourcegraph, and I started working on their repo as a full time job, because it's better to make no money doing something really cool than to halfass something even if you make money. They eventually hired me out of sheer stubbornness and the fact that I wouldn't stop opening PRs and commenting on them (thanks @beyang). I still watched Cursor on the sidelines, because now competing with @cursor was my actual job. There were many truly unconventional decisions from the team that bought them a lot of attacks from VCs and thought leaders. A few examples that come to mind: - The VS Code fork: I have read 100 comments from "smart" people about how maintaining a fork would slow them down since they wouldn't get all the "fixes" from @microsoft. Turns out VS Code hasn't changed one bit in the last 3 years, so I think it was perfectly okay for them to use a fork even if everyone thought it was a stupid thing to do. - Not doing B2B SaaS: The conventional advice for a software company is to approach mega giants, get a massive multi-million dollar contract, and just coast with it. Every acclaimed VC would tell you this, but Cursor didn't do that. They reached roughly $100 million in revenue with no sales team. Many other coding agent companies failed because they wasted 6 months chasing boring enterprises and convincing them to try AI coding agents, when they could have spent that time making something that people actually loved. A lot of Cursor's choices were against startup playbooks. There was a crazy amount of stubborn disagreeableness involved in what made Cursor successful, and if they hadn't made those choices they wouldn't have come this far. All that being said, for the third time in my life I missed an opportunity for generational wealth but I am also really happy that I get to work on one of the most interesting problems of our generation: the "end of software engineering." I worked on so many truly novel things at @sourcegraph and then @cline that I've lost count of how many times I reviewed PRs that were the first of their kind and later became the industry standard for how you're supposed to do things with coding agents, like MCP, tool calling, and different RAG methods. Although I made good money and missed out on generational wealth for now, that doesn't deter me even one bit. I think the only real moat for anything is a genuine love of the game. I don't care if I fail a few more times and have a few more misses on generational wealth, as long as I can confidently say that I gave my heart to the most interesting problem for me. In my opinion, the thing that made Cursor win when other companies had just as many smart people was a genuine love of the game. Not a half-split between "this is what I love in software" and "this is how I make money off it," forever hunting for an iffy compromise between the two. If you love the game, it really doesn't matter how many times you skip the max wealth returns through sheer probability you'll eventually hit a big run. There's no need to be depressed about making less. You should be far more concerned if you don't love the game at all. Because love is the real moat.

I’m convinced at this point that Farza will be the closest we get to a Steve Jobs in this generation.

really proud of the Cursor team - huge congrats to them on the $60 billion exit! throwback to Feb 2022, when I was an early beta user and they reached out about applying to join as employee no.10 if I wasn't building @Cap, this would be a generational fumble 💀

Now it’s official. JetBrains remains the last big independent player in developer tooling. Our job is to deliver the most cost-efficient, deeply integrated, and genuinely enjoyable AI experience across our IDEs and beyond. We’re on it. Stay tuned.

SpaceX acquired Cursor for $60B. A lot of people dismissed Cursor as just a “GPT/Claude wrapper” and a VS Code fork. But it became a company capable of post-training its own models. That is a big reason Elon wanted it. I think more companies will post-train custom models to own the weights and bring model costs down. Congrats to the @cursor_ai team and Elon! Now the real question: will Cursor continue its no-shoes culture inside SpaceX offices?

We did it. Natty is shipping nationwide. Setting a new standard for ice cream. Natty delivers a creamy, indulgent taste, elite macros, and clean ingredients. ELITE MACROS 35g protein, 430-460 cals, 9-15g fiber, and only 9g added sugar (per pint) CLEAN INGREDIENTS No Seed Oils No Artificial Sweeteners/Flavors No Allulose/Sucralose/Erythritol Build your box, while supplies last: nattyicecream.com/pages/buil…

WIP Take - screen recording app what do you think?

I want an app that connects to everything I use (Calendar, Slack, Screen Time, email, etc) Then it tells me exactly what I should be working on, what's distracting me, and where my time is going You could probably one-prompt this with @10x_apps

Pool’s explosive success comes down to one powerful formula: 👉Artistic and cultural symbols 👉Deep technological insight 👉A hyper-common everyday behavior: screenshots They built far more than a screenshot app. They turned digital clutter into personal aesthetics, and into future AI context. 1/ First: artistic and cultural symbols. The game-like interface, beautiful water animations, highly recognizable logo, and the product’s overall visual metaphor give Pool a strong sense of culture. It does not only solve an efficiency problem. It creates emotional and aesthetic identification first. 2/ Second: technological insight. Local CLIP models Gemini intents turn messy screenshots into structured, searchable, and actionable personal memory. The deeper insight here is: Your camera roll is not just a record of the past. It can also become the context through which future AI understands you. 3/ Third: everyday behavior. Everyone takes screenshots. We screenshot beautiful things. We screenshot useful information. We screenshot things we do not have time to process yet. Screenshots are one of the most universal, low-friction, high-frequency behaviors in digital life. 4/ What makes Pool powerful is that it did not try to invent a new behavior. It captured an existing daily action and reframed it as: A more beautiful experience A clearer memory system A more valuable AI context layer 5/ Many apps only get one layer right. Some have aesthetics but lack technological insight. Some have technology but lack cultural symbols. Some capture a behavior but fail to turn it into a new experience. Pool combines all three. 6/ So the biggest takeaway for me is this: The products that win in 2026 may need to answer three questions at once: Does it have cultural symbols? Does it have technological insight? Does it embed itself into a real, frequent, universal everyday behavior? That is a product blueprint worth studying.

Xcode 27 has a markdown editor now and it's amazing!